Compress PDF files the right way: a practical guide

Text in a PDF is remarkably cheap. A page of body text is stored as drawing instructions plus references into an embedded font, and it typically costs a few kilobytes. The full text of a novel fits in the space one photograph occupies. So when a PDF weighs tens of megabytes, the weight is almost never the words.

Scans are the classic offender. A scanner does not produce text; it produces a full-page photograph of text. At 300 DPI, a Letter-size page becomes a 2550 by 3300 pixel image, roughly 25 MB uncompressed in color. JPEG encoding brings that down to 1 to 3 MB per page, which still means a 30-page scan lands anywhere from 30 to 90 MB. Office copiers left on their 600 DPI color default make it four times worse.

Digitally created files have their own bloat patterns. Design tools often embed photographs at full original resolution even when the layout displays them at thumbnail size, and Word documents converted to PDF carry every pasted image at whatever resolution it arrived in. PDFs that have been edited repeatedly can also accumulate dead objects from incremental saves, plus fully embedded fonts where a small subset would do.

PDF compression works on two layers, and the distinction decides how much quality you give up. The lossless layer restructures the file without changing a single pixel: it re-deflates content streams, removes objects nothing references, deduplicates repeated resources, and subsets fonts so only the glyphs actually used ship with the file. This costs nothing visually, but the gains are usually modest, on the order of 5 to 20 percent.

The lossy layer is where files shrink by 70 to 90 percent, and it operates on images in two ways. Downsampling reduces resolution: cutting a 300 DPI scan to 150 DPI quarters the pixel count, because resolution scales in both dimensions at once. Recompression then re-encodes those pixels with stronger JPEG settings, which discard the fine detail the eye is least likely to miss.

JPEG specifically works on 8 by 8 pixel blocks and throws away high-frequency information, which is a technical way of saying it softens sharp transitions. Photographs survive this gracefully. Scanned text does not, because every letter edge is a sharp transition; over-compressed scans grow the familiar halo of ringing artifacts around each character. This is why careful compression treats pure black-and-white scans differently, using bilevel schemes such as CCITT G4 or JBIG2 that are built for text and line art rather than photos.

One reassurance: none of this touches real text. If your PDF contains actual selectable text rather than pictures of text, compression leaves it fully searchable and copyable. Only the image data gets renegotiated.

Compression settings make more sense when you start from the ceiling you need to fit under. Gmail caps attachments at 25 MB and Outlook at 20 MB. Job application portals commonly enforce 2 to 10 MB per file. Government and immigration portals are the strictest, with per-document limits of 1 to 4 MB being routine.

Divide the limit by the page count to get a per-page budget. A 20-page scan that must fit under 4 MB gives you 200 KB per page, which is comfortable at 150 DPI with medium JPEG quality. A 100-page bundle under the same limit gives you 40 KB per page, which means aggressive settings, grayscale conversion, or splitting the file into parts.

As a default, pick the lightest compression that clears your limit. A high-quality setting, roughly 150 DPI with gentle JPEG, suits anything a person will read carefully or print, such as contracts and reports. Maximum compression, 100 DPI or lower with strong JPEG, is for reference copies where on-screen legibility is enough. And always compress from the original file rather than an already-compressed copy; each lossy pass compounds artifacts the way a photocopy of a photocopy degrades.

A handful of situations cover most of the compression jobs people actually run, and each has its own target worth knowing in advance.

Sometimes the better answer is not a compressor. If the PDF was exported from Word, PowerPoint, or InDesign and came out huge, re-export it with image downsampling enabled in the export dialog. The exporter can resample from the original assets, which produces a cleaner result than recompressing the bloated output after the fact.

If the file is a long scan and the limit is tiny, split it into logical parts instead of crushing quality. Two 3 MB files that stay legible beat one 5 MB file that does not, and most portals accept multiple uploads.

And when the images are the point, think twice about lossy compression at all. Blueprints, radiology images, photographic evidence, and print-ready artwork lose exactly the detail they exist to carry. Share those through a file link or a large-transfer service, and keep the original untouched.